In this study, an exploratory analysis was carried out to gain a better understanding of the potential impacts of the two phases (negative and positive) of the Pacific-Japan (PJ) pattern on tropical cyclone (TC) activity affecting the Korean Peninsula (KP) and TC-induced extreme precipitation events over five major river basins in Korea. The results show that large-scale atmospheric environments during the years in the positive PJ phase (referred to as positive PJ years) are more favorable for TC activity than those during the years in the negative PJ phase (referred to as negative PJ years). It is found that wind shear is weaker, rising motion is stronger, and relative humidity is higher over the KP in positive PJ years than in negative PJ years. TCs affecting the KP during positive (negative) PJ years tend to occur more to the southwest (northeast), recurve at locations more to the northwest (northeast), and show an increase (decrease) in frequency over Korea and Japan. As a result, TCs making landfall are found more often over southeastern South Korea during positive PJ years. Despite the relatively modest sample size used in this study, we expect that the results described herein will be useful in developing a critical support system for the effective reduction and mitigation of TC-caused disasters as well as for water supply management in coupled human and natural systems.
 Extreme precipitation events induced by monsoon-related frontal systems and tropical cyclone (TC) activity in East Asia (EA) are critically important not only for their episodic effects such as floods, which have great socioeconomic impacts, but also for their significant contribution to freshwater supplies [Zhou et al., 2006, 2007; Zhou and Chan, 2007; Yuan et al., 2008; Chan and Xu, 2009; Gu et al., 2009; Kim and Jain, 2011; Zhu et al., 2011; Li et al., 2012; Li and Zhou, 2012]. Climate variability in EA is closely related to large-scale atmospheric circulations represented by the Pacific-Japan (PJ) teleconnection pattern during the boreal summer [Nitta and Hu, 1996; Wakabayashi and Kawamura, 2004; Kawamura and Ogasawara, 2006; Choi et al., 2010]. The PJ pattern is associated with convection activities over both the tropical western North Pacific (WNP) and EA regions, which have a significant influence on the weather conditions over EA [Nitta, 1987; Kosaka and Nakamura, 2010]. While the influence of the PJ pattern on the EA summer monsoon has been investigated extensively [Huang and Li, 1987; Tsuyuki and Kurihara, 1989; Kawamura et al., 1996, 1998; Kosaka and Nakamura, 2006], relatively little research has been carried out on the relationship between the PJ pattern and TC activity. Kawamura and Ogasawara found that TCs could alter the vertical structure of the PJ pattern by acting as synoptic-scale convective sources that induce a stationary Rossby wave train over the WNP and result in a transition from a baroclinic to a barotropic structure in midlatitude, characterizing the vertical structure of the PJ. They also pointed out that the east-west pressure gradient at low levels between a TC and an anomalous anticyclone centered over the east of Japan might cause heavy rainfall along Japan's Pacific coast; however, this observation is confined to the month of August.Yamada and Kawamura found that a dynamic link between TCs and the PJ pattern triggers the activity of the Baiu-Changma-Meiyu front in early summer and/or the Akisame front in autumn, bringing in heavy rains to Japan. Previous studies have focused on changes in the PJ pattern due to WNP TC activity and their joint contributions to the potential for heavy summer rainfall. However, seasonal TC activities resulting from the PJ pattern are still not clearly understood. More recently,Choi et al. showed that the PJ pattern has a significant influence on WNP TC activities, but their assessment focused mainly on changes in large-scale atmospheric environments and involved all the extra-tropical cyclones (ETs) over the WNP during the summer season (July, August, and September). In this study, we focused mainly on TCs affecting the KP and the spatial and temporal variability in warm season (June–September) precipitation within the context of two phases of the PJ pattern. Furthermore, the need for improved water information on how and where water managers or stakeholders should adapt their water systems motivates a reappraisal of identifying the relative contribution of typhoon to seasonal and spatial variations in extreme precipitation events over the KP region.
 The purpose of this diagnostic study was (1) to analyze the relative impacts of the PJ phases on TC activities affecting the KP, (2) to explore the main features of the associated large-scale atmospheric environment, and (3) to investigate the seasonal and spatial variations in the TC-induced extreme events for the five major river basins in South Korea. In what follows, data and methods are described inSection 2. The analysis results associated with TC activities due to the PJ pattern are presented in Section 3. Section 4discusses the TC-induced precipitation variability over the KP associated with the two phases of the PJ pattern. Finally, a summary and conclusions are presented inSection 5.
2. Data and Methodology
 Monthly atmospheric data including 850-hPa wind, 500-hPa pressure vertical velocity, and 500-hPa relative humidity were obtained from NCEP-NCAR reanalysis with a spatial resolution of 2.5° longitude by 2.5° latitude [Kistler et al., 2001] for the period 1966–2007. The typhoon track data for the WNP in the period 1966–2007 was obtained from the Typhoon Research Center (TRC) (http://www.typhoon.or.kr, accessed 10 February, 2012) and Japan Meteorological Agency (JMA) (http://www.jma.go.jp/jma/jma-eng/jma-center/rsmc-hp-pub-eg/trackarchives.html/, accessed 10 February, 2012). We also used spatially averaged daily precipitation data (June–September) available from the Korean water resources management information system (Water Management Information System (WAMIS), http://wamis.go.kr/eng/ accessed 10 February, 2012). To understand the WNP TC activity due to the dominant atmospheric circulations, we defined strong positive (negative) PJ patterns as suggested by Wakabayashi and Kawamura . They defined the PJ pattern as the difference in 850-hPa geopotential height anomalies between two grid points to the east of Taiwan (125°E, 22.5°N) and east of Japan (155°E, 35°N) during the boreal summer, as shown in the following equation:
where Z indicates the geopotential height anomalies. Figure 1 shows the normalized summer PJ pattern for the period 1966–2007. Positive (negative) PJ events are defined by the seasonal PJ index exceeding +1σ (−1σ) in the June–September season. Thus, the eight strongest positive PJ events (1972, 1975, 1978, 1979, 1989, 1999, 2000, 2004) and the seven strongest negative PJ events (1967, 1971, 1983, 1991, 1993, 1995, 1996) were selected for the study period of 42 years.
 To estimate the relative contribution of episodic TC events to seasonal precipitation, we conducted an exploratory analysis, as proposed by Kim et al. [2012b], by considering the four-day residence time of a typhoon whose track intersected with a specified Korean domain (120°E–138°E, 32°N–40°N). This restricted domain was used by Korea Meteorological Administration to identify typhoon impacts over the Korean peninsula [Korea Meteorological Administration (KMA), 1996; Kim et al., 2012b]. In the period of 1966–2007, 208 TC events were considered to be KP-affected TCs. Composite analysis with the Student's ttest and Monte Carlo resampling techniques for statistical significance test were conducted mainly to determine the impacts of the two PJ phases on the WNP TC activity, including KP-affected TCs.
3. Characteristics of WNP TC Activity Associated With the PJ Pattern
 Typhoons in the WNP have a strong influence on the hydrometeorological variables over the KP [Kim and Jain, 2011; Kim et al., 2012b]. Recent climate assessments based on high-resolution dynamical models have also indicated that the number and intensity of TCs in the WNP are likely to increase slightly in the 21st century [Knutson et al., 2010; Emanuel et al., 2008]. In particular, Korean river basins are vulnerable to hydrometeorological extremes. Therefore, a better understanding of TC activity may be useful in developing a critical support system for the effective reduction and mitigation of TC-caused disasters. In this section, we analyze the potential impacts of the two PJ phases on TC characteristics, including genesis position, frequency, tracks, intensity, and lifetime, mainly on the basis of KP-affected TCs. Furthermore, the large-scale atmospheric environment associated with these TC activities is discussed insection 3.4.
3.1. TC Genesis Frequency
 The TC genesis is defined as the location where a TC recorded first in the historical data provided by the TRC and JMA. Figure 2shows the difference in the genesis location of TCs that affected the KP during the positive and negative PJ phases over the WNP. The bivariate empirical distribution functions for the genesis positions are shown by contour lines (lower quartile, median, and upper quartile levels are shown). The genesis position of KP-related TCs, centered in the northeastern Philippine Sea, is farther northeast during negative PJ years (average: 142.5°E, 16.8°N) than during positive PJ years (average: 139.2°E, 16.0°N), but the difference between two mean genesis positions of TCs is not statistically significant at the 90% confidence level.
Figure 3 shows the difference in TC frequency between the two PJ phases. According to historical records, 706 TCs were recorded in the WNP during boreal summer (June–September) for the period 1966–2007. For each year, 16.8 TCs occurred in the WNP with a coefficient of variation (CV; the ratio of standard deviation to mean) of 0.24. Further, 208 KP TCs passed through the specified Korean domain (120°E–138°E, 32°N–40°N) during these 42 years. Among these TCs, 5.0 TCs related to the KP occurred in the WNP during summer of each year, with a CV of 0.37. During positive PJ years (Figure 3a), the average number of KP-affected TCs (5.5 TCs/year) was higher than the long-term climatological mean. This difference is statistically significant at the 90% confidence level. On the other hand, during negative PJ years (Figure 3b), the number of KP TCs (4.3 TCs/year) was lower than the climatological mean, which is statistically significant at the 90% confidence level. For all TCs in the WNP, the difference in TC frequency was not statistically significant during positive PJ years (Figure 3a; 17.1 TCs/year), but showed above-normal significant values (17.7 TCs/year) during negative PJ years (Figure 3b).
3.2. TC Track and Recurving Location
 The recurving position of TCs is defined as the location where a TC reaches its westernmost point and shifts from northwestward to northeastward [Choi et al., 2010; Li and Chan, 1999]. Nonrecurving TCs have been excluded from this analysis. TC tracks and recurving location are affected by the meridional and zonal shift of the WNP subtropical high [Choi et al., 2009; Liu and Chan, 2008].
Figure 4ashows the difference in TC activity between positive and negative PJ years for the entire WNP. Here, TC activity is calculated by counting the number of TC occurrences in each 2.5° × 2.5°grid. TC activity with negative values appears along the eastern edge of the Philippine Sea and extends into the South China Sea (SCS), indicating a decrease in TC activity along the westward path in positive PJ years. On the other hand, an obvious increase in TC activity is observed in the specified Korean domain. This is consistent with the position of the WNP subtropical high. The subtropical high is elongated in negative PJ years, whereas it is confined to more eastern and northern regions in positive PJ years. Such an eastward-confined subtropical high favors the recurvature of TCs to the specified Korean domain during positive PJ years while simultaneously leading to a reduction in cyclogenesis in the subtropical high region, as clearly seen inFigure 4b. It is also worth noting that TCs entering the specified Korean domain tend to be stronger during positive PJ years. About 64% of TCs can attain typhoon intensity or above (≥64 knots or ≥33 m/s) in positive PJ years as compared to only 47% in negative PJ years. In other words, more TCs with strong intensity tend to affect Korea during positive PJ years. Especially, during negative PJ years, it shows statistically significant reduction in the number of TCs with maximum wind speed exceeding 64 knots (≥33 m/s), as a result of decrease in the intensity of KP-affected TCs, significantly suppressed precipitation modulated from TCs is predominant over the KP (seesection 4 for details).
 On the basis of the above definition of a TC recurving location, we found that 77.3% (34 TCs) and 93.3% (28 TCs) of KP-related TCs occurred in positive and negative PJ years, respectively.Figure 5shows the difference in TC recurving location between the two PJ phases, shown by the bivariate empirical distribution with contour lines (lower quartile, median, and upper quartile levels are shown). In this figure, the recurving location of KP-related TCs is found farther to the northeast in negative PJ years (average: 124.6°E, 27.6°N) than in positive PJ years (average: 127.3°E, 28.2°N), but their difference in the mean recurving location of TCs is not statistically significant at the 90% confidence level.
3.3. TC Lifespan
 The lifespan of a TC is defined as the total number of days of KP-affected TCs recorded by the TRC and JMA (Figure 6a). The lifespan of a potential KP-related TC is defined as the period from TC genesis to the date at which the typhoon track intersects with the specified Korean domain (Figure 6b). In Figure 6, the empirical distribution of TC lifetime is shown in the form of a violin plot [Hintze and Nelson, 1998], which combines a box plot and a density trace to display the structure found within the data.
 During positive PJ years, TCs pass through the KP domain (−0.08 day) and have a slightly shorter lifetime than the climatological mean value. On the other hand, TCs during negative PJ years (0.31 day) have a longer lifetime than the climatological mean value (Figure 6a), but their difference is not statistically significant. Figure 6b shows that the approach time of TCs to the KP domain during positive PJ years (0.24 day) is longer than that during negative PJ years (−0.18 day). This implies that TCs formed during positive PJ years spend a longer time over the ocean, allowing them to become more intense when they approach to KP domain.
3.4. Large-Scale Atmospheric Environments
 In order to explain the variation in TC activity associated with different PJ patterns, different TC-related large-scale parameters, as suggested byGray , are investigated in this section. Figure 7 shows the anomalies in total vertical wind shear in positive and negative PJ years. A prominent difference is the significantly weaker (stronger) wind shear at the midlatitude of 30°–40°N, which coincides with the Korean domain in positive (negative) PJ years. Because strong wind shear tends to destroy the vertical structure of TCs, the weak shear in the Korean domain during positive PJ years favors the subsequent development of TCs, thus explaining why more TCs enter the Korean domain and have a higher chance of intensification in positive PJ years. This is consistent with the observation made of Choi et al.  that TCs during positive PJ years tend to be stronger than those during negative PJ years. Apart from wind shear, an anomalous rising (descending) motion (Figure 8), together with high (low) relative humidity (Figure 9), as well as low-level convergence (divergence) anomalies (Figure 10), can also be observed in the Korean domain during positive (negative) PJ years. The anomalous rising (descending) motion is associated with the cyclonic (anticyclonic) circulation centered to the northwest of Taiwan during positive (negative) PJ years, providing a favorable environment for the subsequent passage of TCs. It is also worth mentioning that the composite results here are generally unaffected by the removal of the TC-related days. This suggests that the difference in atmospheric environmental parameters under different PJ conditions is not a response to TCs. Rather, they serve as contributing factors that lead to significant differences in TC activity in the Korean region.
 In the following sections, we investigate the impacts of the two phases of the PJ pattern on seasonal precipitation totals and the frequency of heavy precipitation events coincident with TCs at the regional scale (112 sub-basins in Korea), where adaptation and migration can occur.
4. TC-Induced Precipitation Variability Associated With PJ Phases
 During the study period, approximately two-thirds of the average annual precipitation (ranging from 821 mm to 934 mm in five major river basins) in Korea occurred during the boreal summer (June–September). During this warm season, episodic events stemming from TCs, which have the potential to create flood hazards [Kim et al., 2006; Chang et al., 2009] and impact seasonal water supplies [Kim and Jain, 2011], account for nearly a quarter of the seasonal precipitation totals, with high spatial variations due to topographic effects. Kim and Jain  provided detailed information on the precipitation pattern for the five major river basins in South Korea (also refer to Figure S1 in auxiliary material). In the present study, an episodic TC-based approach [Kim et al., 2012b] was used to separate the precipitation or their remnants recorded in the specified Korean domain, which was used to identify KP-affected TCs over the WNP [KMA, 1996; Kim et al., 2012a, 2012b].
4.1. TC-Related Seasonal Precipitation Totals
Figure 11shows composite analysis results for TC-induced seasonal precipitation totals associated with the PJ teleconnection pattern. The regional pattern of TC-induced precipitation anomalies shows a remarkable difference between positive and negative PJ years. For the positive PJ years (Figure 11a), positive precipitation anomalies are observed to the northwest of the Han River basin (51.5% of the total area) and in the southern KP (Nakdong River basin: 79.0%, Sumjin River basin: 89.3%, Youngsan River basin: 91.3% of the total area). The western river basins (especially, the Geum River basin, 71.6% of the total area) and the eastern and southern parts of the Han River basin (48.5% of the total area) are indicated by the negative pattern of TC-induced precipitation anomalies; however, these anomalies are not statistically significant. The most significant changes in the total precipitation coincident with TCs are found in the Nakdong River basin, which shows the largest fraction of seasonal precipitation (26.1%) induced by TCs from June to September. Five of the 33 sub-basins (17.2% of the total area) in the Nakdong River basin show a statistically significant precipitation anomaly pattern with positive anomalies.
 In contrast, the spatial anomaly patterns of TC-induced precipitation during negative PJ years are very different from those during positive PJ years. For negative PJ years (Figure 11b), negative precipitation anomalies coincident with TCs are observed in all the sub-basins of South Korea. The following river basins show a statistically significant precipitation pattern with negative anomalies: Han River basin (55.1% of the total area, 15 out of 29 sub-basins), Nakdong River basin (40.6% of the total area, 12 out of 33 sub-basins), and Geum River basin (6.6% of the total area, 1 out of 21 sub-basins). The Sumjin and Youngsan River basins show statistically significant negative anomalies in all the sub-basins. As a result, abundant precipitation induced by TCs is found in southern South Korea during positive PJ years. On the other hand, significant suppressed precipitation stemming from TCs occurs over the KP during negative PJ years. These distinct changes in precipitation during the two PJ phases can also be seen in the seasonal Global Precipitation Climatology Project (GPCP) precipitation (refer to Figure S2 inSupplementary Information).
4.2. TC-Related Heavy Rain Days (≥50 mm/day)
 The frequency of heavy rain days (≥50 mm/day) is highly correlated with boreal summer rainfall over the 112 sub-basins in South Korea [Kim and Jain, 2011] and threshold of 50 mm/day was reported as approximately the 90th percentile of warm season precipitation totals for major stations over South Korea [Chang and Kwon, 2007].
Figure 12shows the composite analysis for the number of days on which the TC-induced daily precipitation exceeded 50 mm. This result is similar to that inFigure 10. For positive PJ years (Figure 12a), above-normal anomalies in heavy rain days are found in southern South Korea. In the Nakdong River basin, 4 out of 33 sub-basins, which account for 8.3% of the total area, show significant positive anomalies for heavy rain days. However, the western and interior regions of South Korea, including the Geum and Han River basins (24.9% and 5.6% of the total area, respectively), as well as the northern and eastern coastal regions of the Nakdong river basins (9.1% of the total area), show significantly below-normal anomalies for heavy rain days. In contrast, for negative PJ years (Figure 12b), negative anomalies are dominant over the KP, except for several sub-basins in the Nakdong and Geum River basins. The most significant negative anomalies are found in the Sumjin (15 out of 15 sub-basins) and Youngsan River basins (13 out of 14 sub-basins).
5. Summary and Conclusions
 In this diagnostic study, an exploratory analysis was carried out to better understand the potential impacts of the two PJ phases on TC activities affecting the KP and TC-induced precipitation variability over Korea's five major river basins. Our results can be summarized briefly as follows:
 1. It is found that the genesis position of TCs affecting the KP is more likely to appear father to the northeast during negative PJ years than during positive PJ years. For each positive PJ year, 5.5 TCs occur, which is significant higher than the climatological mean value. On the other hand, fewer TCs (4.3 TCs/year) occur during negative PJ years.
 2. TC activities associated with the PJ pattern are found to be closely related to large-scale environmental parameters. During positive PJ years, a higher number of TCs enters the Korean domain. This can be attributed to weak wind shear, strong rising motion, and high relative humidity in the Korean region, factors which provide favorable conditions for the subsequent passage of TCs. In addition, the eastward and northward shift of the WNP subtropical high during positive PJ years favors the recurvature of TCs, whereas the elongated subtropical high during negative PJ years does not. Consistent with all these large-scale parameters, the number of TCs affecting the Korean region is higher in positive PJ years than in negative PJ years.
 3. The spatial distribution of TC-induced precipitation shows a remarkable difference between the two phases of the PJ teleconnection pattern. During positive PJ years, TC rainfall is significantly above normal in the Nakdong River basin (17.2% of the total area). In contrast, during negative PJ years, significantly suppressed precipitation modulated from TCs is observed over the KP. This result is also evident from the frequency of heavy precipitation. The Nakdong River basin (8.3% of the total area) shows significant positive anomalies in the number of heavy rain days associated with TCs during positive PJ years. In contrast, during negative PJ years, significant negative anomaly patterns in heavy rain days are found in southern South Korea. The above results clearly indicate that different TC-related parameters have different effects during positive and negative PJ years.
 The analysis results presented here reveal characteristics of extreme events caused by the process of TC-related moisture transport associated with the PJ teleconnection pattern. Thus, information about the systematic variability associated with atmospheric teleconnection can be incorporated into adaptation strategies and risk management of water resources in coupled human and natural systems. We believe that despite the relatively modest sample size used in this study, our results have the potential for providing a better understanding of the effects of the PJ pattern on TC activities and the spatial and temporal variability in summertime precipitation at the sub-watershed scale, where adaptation and mitigation occur over the KP. Future work based on numerical simulations is also required to examine the seasonal evolution of teleconnection patterns, their interactions with typhoon and tropical moisture plumes in the atmosphere.
 This research is supported by Joint National Natural Science Foundation of China Project 41175079, and the City University of Hong Kong Strategic Research grants 7002717 and 7002780. The authors are thankful to Johnny Chan for his insightful comments and suggestions. We appreciate the constructive input from the Editor and three anonymous reviewers.